[sesqu]

Some of the data seem a little suspect. The grade correlation result is unreasonably strong, and the bin sizes aren't specified, among other shortcomings. That said, this was a very entertaining read.

edit: Appears to be from 1990.

[gus_massa]

Exactly. I was reading skeptically and trying to guess what were the errors in the methodology that create this effect. Until I see the Figure I, with the almost aligned dots (If you ever made some physics or statistical experiments, you would be happy to get a measure like this.) The dots are almost aligned and one of the axes is the A/B/C/D/E grade, that is very indirect nonlinear measure. A graph like that usually means that the data are cooked.

Apparently this is only a humor piece, but many of the comments seams to take this seriously.